Re: New lossless image compression algorithm

Well, it's a bit of risk for us to retest an algorithm in
developement all
over again. Would be better to have a semi-final version, or at least
a stable fork available.

Well the algorithm should be quite stable, I've tested the actual

Well that too, but I mean stable to changes. Do it like D'exupery
said:
"Perfectionism is reached, when you can't take anymore away", I think
it
was him. The thing is, that the algorithms should comply with the goal
to actually use it's files: "I save them, I use them, with version
0.234a I
don't want to loose my files".
Or you convince me that there is something special with it, that it
re-
presents a set of algorithm not covered at all (well this in relation
to
my db). I don't have all 579 different wavelets tested because of
that.

BTW: You talk in the algorithm description about the special spatial
filtering-step. The color-correlation thingy is normaly called "inter-
band
prediction", and is an old hat. Here are some excellent sources:

"Multispectral Image Coding", Daniel Tretter
"Lossless Compression and Interpolation for High Quality Still Images
and Video Sequences", Stefano Andriani

And this coder uses it since 1995:

"Adaptive Prediction Trees for Image Compression", John Robinson

The creators of CALIC also wrote a paper about extending it with a
inter-plane context, don't see the title of the paper in this moment.

version on many images (surely not on GBytes though !) and it seems to
work correctly; it also embeds an hash code in the compressed file to
check if the original image is rebuilt correctly. I would be very
grateful to any third party who wants to test it also because this
could be another test for its stability, but I surely understand the
risk.

This sounds as if you are talking about Lossless-JPEG. There is
another
proposal LOCO, which is extremely hard to beat. I'm actually not avare
that
Lossless-JPEG is used anywhere, whereas LOCO allready is in cameras
and on Mars btw.

Nope, I really used the LOCO algorithm (downloaded fromhttp://www.hpl.hp.com/loco/).
I too passed some time to understand the problem since I knew of the
high performance of the program, but it seems that the current
implementation, even in the version for 24-bit images, does not apply
a color transform... maybe programs used in cameras and imaging from
mars included it ? Any explaination is welcome.

Wellll, I would group the failings of LOCO into two categories:
histogram-sparseness and pre-filtered images, ah and small images.
LOCO gets the boost beyond 6MPixel. You just don't have to expect LOCO
to be the best allways, but I would say LOCO is very persistant in
being near the best, which makes it attractive in a variety of ways.
This adds up to, that a minimal LOCO requires around 24k total memory,
and the number of ops is I don't know maybe 250 per pixel (just a
rough qiuck guess).
I don't know if the Mars-Rover uses the YCbCr RCT of JPEG-2000, but
it is a possibility as they also use a cousin Wavelet-System for lossy
compression. I doubt they use inter-band correction of prediction
residuals. Context get's big fast, memory-requirement tripples,
algorithm complexity too if you consider coding planes in distinct
resolutions (let's hail to Mr. Bayer, we just love 4:2:2 RGB, it's so
much more high quality than 4:2:2 YCbCr from 1980!).

All the best,
Stefano

To you too [reading your webpage ...]
Niels
.

Relevant Pages

  • Re: How much lossless compression is possible in images?
    ... >So how much compression could you get? ... compression that halts early, given that the ideal algorithm for that ... on half of the images (99.99999999999% of 4GB images ... the data will just be a UTM copy algorithm with the 4GB input string ...
    (sci.math)
  • Re: New lossless image compression algorithm
    ... which has a low-complexity ultra-fast compression technique. ... explanation of how the algorithm works. ... this was mostly optimized towards photographic images and not computer generated images. ... I would furthermore consider a more modern entropy coding backend, ...
    (comp.compression)
  • Re: Prediction in differencing pixels with the left neighbour.
    ... at the moment I am working on lossless compression. ... I would prefer on a simple, clear and fast algorithm ... I can compress images near Photoshop HP JPEG-LS plugin can do ... as sayid before I compute for each pixel the difference with the ...
    (comp.compression)
  • Re: How much lossless compression is possible in images?
    ... The point here is any compression algorithm, ... operating on the total set of inputs, ... The average file length of the original 16^1,000,000 images is also ... Finding a particular compression algorithm for that image doesn't help ...
    (sci.math)
  • Re: faster rendering of eps pictures in YAP?
    ... Images on the hard drive take more time to load than images ... by the cpu to implement algorithms, ... So using a more efficient algorithm saves time. ... Lower resolution and smaller images will take less time, but then quality will be sacrificed. ...
    (comp.text.tex)